Biomedical Engineering Reference
In-Depth Information
shoulder region of the breast, which can deform to a greater extent compared to the
other regions of the muscle. These outer areas are still needed to be included during
identification to avoid introducing possible bias by arbitrarily including only
specific regions of muscle. The maximum error observed in the muscle provides
another indication of the improvement of using the heterogeneous versus the
homogeneous models (4.5 mm max error). Based on these results, we conclude
that the mechanical effects of the pectoral muscle should be accounted for in
biomechanical breast models subject to gravity loading in order to accurately
simulate the deformation in the shoulder region of the breast.
To compare the neo-Hookean stiffness parameters (
c
) identified for the different
tissues in this study against shear moduli (
m
)
reported in literature, we have used the
following relationship;
m
¼
2c [
15
]. The estimated shear modulus of the breast
tissues (
0.1 kPa, which we believed largely represented the
modulus of fat due to the relatively large ratio of fat to fibroglandular tissue) is
below the range of values reported in the literature for ex-vivo samples of breast
tissue subject to mechanical testing and elastography (
m
breast-tissue-identified
¼
25 kPa)
[
16
]. This discrepancy may be due to the fat tissue being in a more liquefied (and
thus softer) state at body temperature compared to room temperature [
16
,
17
].
The estimated shear modulus of the muscle (
m
fat-literature
¼
0.5
0.52 kPa) was within
the range of reported transverse shear moduli for human skeletal muscles in static
body postures (
m
muscle-identified
¼
1.2 kPa) [
16
].
While the method of segmentation implemented in this paper provides an
accurate surface profile of the pectoral muscles, it does involve manual intervention
to select threshold levels. However in a clinical application, this would not be
practical and a fully automatic method for segmentation of the muscle surface
would be required which would be capable of attaining the same level of accuracy.
We are currently investigating the use of a multi-atlas segmentation process for this
purpose [
18
], since the location and makeup of the muscle is fairly consistent across
subjects lending it well to these approaches.
Fibroglandular tissue was not separately accounted for in this study since its
distribution was interspersed throughout the fat tissue in the central regions of the
breast. Therefore, it was unclear whether the necessary information for identifying
its mechanical properties could be obtained using only data segmented from the
reference supine MR images. To address this shortcoming, one possibility would
be to identify the parameters by making more direct use of the medical images in
the estimation procedure. For example, the objective function could be an image-
based similarity measure (such as normalised cross-correlation [
19
]) between the
embedded and model-warped prone MRI and the reference supine MRI.
One of the main challenges in modelling the mechanical behaviour of soft tissue
is the accurate determination of the constitutive properties of the tissue. In this
study, we identified the mechanical properties of the breast tissue using information
extracted from reference supine MR images. This is, however, not an ideal solution
since obtaining supine MR images represents an additional cost and furthermore,
such images are not commonly acquired in clinical practice. Therefore, we are
developing other methods of identifying mechanical properties by capturing
m
muscle-literature
¼
0.25
